Wind driven generator fault identification method based on hybrid neural network

Abstract

The invention discloses a wind driven generator fault identification method based on a hybrid neural network. The method specifically comprises the steps that time domain waveform data of a gearbox are collected, original sample data are built, and the data are labeled; the minimum value of the amplitude, the vibration speed and the kurtosis index in the waveform data are extracted as features; the extracted fault and normal feature values are input into a hybrid network 1D-CNN_Bi-GRU, the hybrid network is connected with a 1D-CNN and a Bi-GRU in series, firstly, the 1D-CNN is used as a primary network to extract sequence local features, then the output of the 1D-CNN is used as the input of the Bi-GRU, the characteristics of the Bi-GRU are utilized, accumulated dependency information from the past in the forward direction and accumulated dependency information from the future in the reverse direction are obtained at the same time, and long-term dependence characteristics of the sequence are further extracted to carry out fault diagnosis; and a model is stored, to-be-analyzed data are input into the model, and a fault classification result is output.

Description

technical field [0001] The invention belongs to the technical field of fault identification for wind power generation, and relates to a key technical method for fault identification of a wind power generator based on a hybrid neural network. Background technique [0002] The field of wind power has developed rapidly in recent years, but the technology in the manufacture and maintenance of related equipment is still immature, and because the installation sites of wind power equipment are generally in relatively harsh environments, how to ensure that the wind power It has become a hot spot for technicians to be able to judge the hidden dangers of failures in advance through prediction. As the scale and cost of a single fan increase, the cost of maintenance also increases significantly. According to the results of the data, the service life of the general fan is about 20 years, and the daily maintenance and repair expenditure of the fan accounts for 10-15% of the total expendi...

AI Technical Summary

Problems solved by technology

[0012] Aiming at the problems of existing wind turbine fault diagnosis methods, such as difficult model establishment, general feature extraction ability, poor generalization ability, low precision and small data, a high-performance method based on bidirectional gated recurrent unit and one-dimensional convolutional neural network is proposed. Hybrid Neural Network Fault Diagnosis Method

Images